Apple's M5 chip has a 16-core Neural Engine that barely gets mentioned in reviews. But Apple has been making it bigger every generation since 2017. Other chip companies are doing the exact same thing. When every major silicon maker starts dedicating more transistors to the same component, that's not a coincidence. You can read their bet in the chip itself.

This video traces the evolution of Apple's Neural Engine from its first 2-core version in the iPhone X all the way to the M5 MacBook Pro. Eight years of investment, generation after generation. Then I look at Qualcomm's Snapdragon X2 Elite and Intel's Lunar Lake and show you the same pattern playing out across the entire industry. Three companies, three different business models, all converging on dedicated AI hardware.

I also break down why Apple's approach to chip design is structurally different from everyone else in the PC space. They design the chip, the memory, the operating system, and the developer frameworks. One company, full vertical. Nobody else controls that much of the stack, and it changes what's architecturally possible. The video ends with the real question underneath all of this: the split between cloud AI and on-device AI, and what that means if you build software.

Tiff In Tech

Apple's Neural Engine has grown from 2 cores in the A11 (2017) to a 16-core engine with GPU-embedded neural accelerators in the M5, representing an 8-year hardware thesis on on-device AI. Apple's vertically integrated stack — from transistors to Core ML — enables unified memory, coordinated workload routing, and private local inference that cloud-only approaches can't match. The gap between what the neural engine can do today and what developers have actually built for it represents a significant opportunity. Qualcomm and Intel are making similar silicon bets, signaling industry-wide conviction that meaningful AI compute belongs on the device, not just in the cloud.

What Apple’s Neural Engine Tells You About the Next Decade